How to structure architecture specific NuGet packages when no AnyCPU assembly exists?

[Danielku15]

Hi everyone.

I have a C++/CLI (aka. Managed C++) wrapper for one of our project libraries and I want to have a NuGet package for that one. I followed the guide for Architecture-specific folders and the packaging works. But as indicated by the documentation:

These assemblies will only be available at runtime, so if you want to provide the corresponding compile time assembly as well then have AnyCPU assembly in /ref/{tfm} folder.

But what do you do if you do not have an AnyCPU specific assembly. You cannot have AnyCPU in Managed C++. In the project where I consume the library I have set the Platform to x64 but still do not have any compile time support for this library.

My structure is :

\runtimes
    \win10-x64
        \native
            \boost###.dll
        \lib
            \net5.0
                \MyLib.dll
            \net48
                \MyLib.dll

As a workaround I will likely place the files from runtimes\win10-x64\lib directly at lib\. Luckily I only need x64 so far but it still feels wrong. Do I need to duplicate my whole library surface into an unused AnyCPU version for compile time support?

I appreciate any help on this topic.

Greetings
Daniel